Almost every other SaaS vendor supports multi-region active-active and Cognito does not.
Who are we talking about here? Google and Azure?
https://support.okta.com/help/s/article/overview-of-enhanced...
Note:
1. Customers with existing user pools created on or before 10:00am Pacific Time, November 22, 2024 will continue having a free tier of first 50,000 MAUs. Advanced Security Features (ASF) will continue to be priced separately and will not have a free tier, just like it has been priced previously.
2. Additionally, customers are eligible to create new user pools with Lite tier in their existing accounts and count those MAUs against the free tier of first 50,000 MAUs. To be eligible, customers' accounts must have had at least 1 MAU in the last 12 months on or before 10:00am Pacific Time, November 22, 2024.
I agree, the author is totally confused.
Lite is equivalent was available before, and there's been a lot of improvements to essentials and plus (passkeys, more customization options). It is a bit weird that you have to be in the Plus plan to export user activity logs, but pricing is difficult. Depending on the number of users you had, lite can cost quite a bit more (5x for 60k users).
The simple pricing, which was one of the key benefits of Cognito, has become more complex. I put together a spreadsheet to show the price changes here: https://docs.google.com/spreadsheets/d/1Nm5BUOjFlqqvaeDTERJm...
It's pretty clear that while there are increased features, if all you need is login, the pricing has increased significantly if you are above 10k MAU. If you are under 10k users, it's a wash, of course.
This spreadsheet ignores quota increase, SAML users, and M2M tokens, which are all charged separately.
However, you can see the plus tier, which includes what was previously "Advanced security features" has gotten significantly cheaper (pricing for that feature sourced from https://medium.com/@demandapi/aws-cognito-advantages-pricing... ). So bravo to them for lowering that.
There's also complexity in switching tiers, since if you switch from Plus to Lite in the middle of the month, you'll be charged Plus for all users who logged in before the switch and Lite for all after. That's incidental, but still added complexity.
And, as a sibling comment notes, there's legacy pricing for anyone who has a user pool with at least 1 MAU in the last 12 months. This legacy pricing lasts until Nov 30, 2025, which gives folks a chance to migrate or adjust business models.
Oh, hello Purdue Pharma!
authentication is critical, you shouldn't be outsourcing this stuff anyhow. learn how to harden your box, use cloudflare tunnel and dont store passwords in plaintext.
its really not hard to do and constantly being gaslighted into paying someone to do it for you because everybody else is doing it is just irresponsible.
I guess one clear difference is the lack of a marketing department from something well funded. I recall another HN comment here that said the best business model is to take something people can do already and mark it up by selling the pain points, that could be whats also helping all these auth as a service vendors.
Go with a proven, vetted, and trusted open source solution.
I think it shouldn't be too hard. I could even add Remix support for you if you wanted to do a contract (I am not able to do major open source work for free right now.)
- Cognito: bad
- Clerk: ok for small scale applications but they're a small company 'moving fast and breaking things'. It's not stable enough for a enterprise grade product that needs robustness.
- Auth0: Good but can get expensive
- WorkOS: Good for B2B, especially if 'directory syncing' is important for your usecase
Unless you would like your data to egress from an AWS datacenter, in which case they are a very, very long way above market value.
> two great products
RDS is also pretty great, and KMS is a pretty good way to store a private key per environment.
Even then, you give some of the discount back as AWS Enterprise Support charges :)
(That majority of the customers might stand for less than the turnover of the minority that enjoys discounts. But that does not help you if you belong to the majority.)
S3? Laughs in egress costs.
AWS considered harmful.
Compared to Hetzner? Come on.
Amazon just prices S3 and EC2 at not-insane rates because they shadow charge you for I/O and network traffic at 10x a competitive rate, things that people don't actually look at when evaluating cloud providers.
SQS is great too. To many people it's reliable and durable, and implements a pretty robust competitive consumer pattern.
https://aws.amazon.com/blogs/aws/improve-your-app-authentica...
I don’t like AWS but god damn they are good marketers and had some good leadership that actually was ahead of the curve. Instead of min/maxing the quarterly earning calls.
Convince a nepo C-level executive of your offerings, wave your massive AWS dick while presenting your deck, throw in a few credits, keep it “cheap” for a number of years. Once the competition fizzes out, or you buy them up. Then nix those teaser rates and jack it up 100X over a decade.
Now AWS is pumping for the next millenia.
Their stock is bumping along at $200. If they can keep people coming in and staying, then the stock can go brrrr for decades. But if they cant, eg the trickle of CTOs repatriating workloads to prem becomes a roar, it wont, and AWS will turn into IBM.
You clearly have strong opinions on how AWS operates, but their stock holders are happy bunnies. What's your prediction?
The question is if we're living in the new 1970s or 2000s.
The auth service space is so strange. Almost every vendor is ridiculously expensive for any B2C use-case. Cognito, with its free tier of 50k MAU, was one of the few relatively cheap options. Even the "open source core" offerings in the space are crazy if you use their hosted version. And their self-hosted versions inevitably end up requiring you to run Postgres, Redis cluster, a background job running task, etc. If you're not getting Cognito for cheap, you're better off just using libraries to roll your own auth service/module instead of going for any off the shelf auth SaaS or self-hosted solution.
AWS team\org business priorities, like P&L computation, changed pretty drastically in 2022-23.
Historically there were a lot of services built, run, and measured on the idea of solving customers diverse needs and making AWS a better place to run your business. This isnt a “loss leader” per se, but 1) profitability may not be your highest business prioirty 2) customers & shareholders valued growth & diversity of offerings above almost everything else 3) business units would set forward looking pricing based on marginal rates 2-3 years out under better utilization models 4) services would not-uncommonoly use “attribution” or “flow through” revenue models for P&L. Eg autoscaling doesnt have a meaningfully price, but it drives (hypothetically) 3% of EC2 instance hours. Autoscaling than books a portion of the 3% of instance hour revenue to their profit center. Cognito (or Route 53 or SSM where I worked) would use this sort of P&L model.
Circa 2022 AMZn shareholders, amazon execs, and the market more broadly turned to Revenue and Profit as the goal, no longer growth per se. This drastically changed a lot of internal business models. No more “free rides”, book revenue, define and execute a plan to be a many million dollar direct revenue business, the old “growth and better together” story wasnt selling.
And i dont think thats a bad thing per se, as a shareholder. I appreciate the focus on proving your value via pricing and usage. But there will be some sad “abandonware” and service shutdowns over the next few years.
Step 2 jack up the prices
"The greatest trick aws ever played was convincing engineers that rolling your own infra is bad and scary."